THE RELATIONSHIP BETWEEN WAGES AND PRODUCTIVITY - USC

International Journal of Applied Econometrics and Quantitative Studies Vol.5-1 (2008)

THE RELATIONSHIP BETWEEN WAGES AND PRODUCTIVITY: TAR UNIT ROOT AND TAR COINTEGRATION APPROACH BILDIRICI, Melike * ALP, Elçin Aykaç Abstract We analyzed the relationship between wages and productivity ni 1990 - 2007 period in Turkey. At the first stage we followed the traditional unit root tests and apply the analysis followed by unit root test procedure proposed by Caner and Hansen (2001). Then we discussed the long run nonlinear relationship between wages and productivity by employing the TAR cointegration analysis developed by Hansen and Seo (2002). JEL Classification: C32, J31 Keywords: wage, productivity, non linear, TAR unit root, TAR cointegration. 1. Introduction In the US, the founder of the approach that wages be based on increases in productivity, the firms have considered productivity levels as the measure of wage rises, while the firms in England have adopted the practice of delivering stocks in return for productivity. The reflection of these instances on labor literature is raising wages after observing increases in productivity or the practice of prior raises in wages in order to increase productivity. The second approach that has been intensely debated after 1980 and known as effective wage approach, has been developed by the European branch of new Keynesians and been used in explaining the reason why wage levels differ from the Walrasian equilibrium values. The studies are not limited to these and the relation between wages and efficiency has been analyzed in different dimensions. Salop (1979), Shapiro and Stiglitz (1984), Hall (1986), Alexander (1993), Hadroyiannis (1997), Dibooglu and Enders (2001), Marcellino and *

Prof. Dr. Melike Bildirici, Yildiz Technical University, Faculty of Economic and Administrative Sciences, Istanbul, Turkey, email:[email protected], and Res. Assist. Elçin Aykaç ALP at the same university, e-mail:[email protected]


Mizon (2001), Welfe andMajsterek (2002), Brügeman (2004), Wakeford (2004) and Christopoulos and Tsionas (2005) have used panel unit root and cointegration analysis in their studies that investigate the relationship between wages and productivity with an approach that also accounts for unemployment as a variable. Bildirici (2004, 2005) has emphasized that there existed no relation between productivity and wages in Turkey. We also accept that peculiar circumstances after 1980 have caused deviations from productivity on a manufactural basis and the relation between labor productivity and wages1 . However, these factors not only caused deviations but also have led the labor productivity to keep as one of the lowest among OECD countries. Consequently, the rate of increase in labor productivity for the period 1980-2000 is merely 2.7% while the same ratio for East and South Asian countries is 5.6% (OECD, IMD). Considering the period 2003-2005, the years by which Turkey has partially gotten rid of the heaviest crisis it has experienced, it still seems to be among the most unsuccessful performers in productivity increases among OECD countries (see figure 1 ). 600 500 400 300 200 100

2003

2005

Figure 1. Labor productivity in OECD countries fort he year 2005 (Purchasing power parity, value added perm an hour) 1

Productivity on manufactural basis is calculated in conformity with the traditional method that requires the output obtained as the result of production be divided by he amount of input. If this value happens to be less than 1, this means that costs are rising. The most important reason for the increase in costs is the direct and indirect taxes and other liabilities incurred by the employer on behalf of the employee. In terms of costs, specifically labor costs will e considered since the most significant item in costs is this one.

94

G ree ce

Tu rk ey

Slo va kia

Po rtu ga l

Po lan d

M ex ico

Ko rea

Hu ng ar y

Ja pa n

Ita ly

Sp ain

Irla nd

En gla nd

Ho lan d

Fr an ce

De nm ark

R ep ub lic

C ze ch

AB D

G er m an y

0

Bilidirici, M., Alp, E.A. The Relationships between Wages and Productivity

Although there exists a productivity gap of 3.8 times (%283) between the most productive country, Ireland and Turkey, the differential in gross wages is merely 67%. Taxes on wages (direct taxes, indirect taxes and other liabilities) are an important reason for the gross wages being high in Turkey. Tax burden on the wages raise gross wages while they break the link between productivity and wage level. In economic literature, the effect of taxes on wages (especially on effective wage) has been discussed and Economists have come to a conclusion that there exists a strong effect. The literature on taxation in labor markets in context of efficiency wage find that the specification of taxes influences wages and employment [e.g. Malcomson and Sartor, 1987; Hoel, 1990; Delipalla and Sanfey, 2001, (see for efficiency wage, Akerlof and Yellen;1986, Shapiro and Stiglitz, (1984), Johnson and Layard, (1986),. Pisauro (1991) Rasmusen (1998) Faria (2001), Hoon and Phelps (1992), Lin and Lai (1994) and Faria (2000). These studies are based on direct taxes as a consequence of the structure in the particular country subject to the study. On the other hand, direct taxes, indirect taxes and liabilities are significant in Turkey2 and payments made by the employer on behalf of the employee (employee’s cost to the firm) increase the gross wages and consequently overall input costs. In 2000,Turkey has been the third among 30 OECD countries in employment taxes (including income tax paid by the employer on behalf of the employee, dole payments, and all employee and employer payments made to the Social Security Foundation). By the year 2001, Turkey has become the first and kept this position in 2004 (calculated from the data by OECD, TISK and IMF). Figure 2 depicts the share of employment taxes in employers’ costs for the period 2001-2005. In this comparison, some indirect taxes peculiar to Turkey are overlooked in order to construct a linkage between Turkey and other countries. 2

Additionally, costly employment and firing practices, application of lump-sum wage increases in collective bargaining procedures, resistance to job evaluation systems, degeneration of the system and application of minimum wage laws are other important factors that lie beneath low productivity increases. The concept of “employee’s cost to the firm” which seems more important than other factors will be treated separately.

95

International Journal of Applied Econometrics and Quantitative Studies Vol.5-1 (2008) 50 45 40 35 30 25 20 15 10 5 Ire lan d

AB D

Ice lan d

Ze ala nd

Ne w

Ko rea

Me xic o

Ja pa n

Ca na da

No rw ay

2001

En gla nd

Ita ly

2003

Au str ali a

2005

Sp ain

Au str ia

Gre ec e

Ge rm an y

Be lgiu m

Hu ng ary

Po lan d

Fra nc e

Tu rke y

Sw ed en

0

Figure 2. Share of Employment Taxes in Labor Cost, (The ratio of the sum of taxes levied on wages, social security premia of the employer and the employee to the overall labor cost (%) including familial tax subsidies) Source: OECD, Taxing Wages Statistics, various years; TISK, Working Statistics and Labor Cost, various years.

With the effect of mentioned factors, productivite seems to be following a path independent from wage. On the other hand, this does not exactly reflect the facts. The relation between productivity and the wage received by the worker must be analyzed. For this reason, we shall consider the wage exclusive of indirect taxes and other payments. Besides, in order to avoid the high cost of employment and to prevent productivity increases falling below wage increases, firms recourse to some practices such as informal employment 3 , foreign and illegal employment 4 , non-primary earnings.5 Such practices generate increases in production but this 3

Even though there are fluctuations between the years 1989-2006, 50% of overall employment is informal. Concentration of employment on agrarian sector has an important role on this outcome. 26.7% of those who are employed in non-agrarian sectors work at informal sectors (www.tuik.gov.tr) 4 In the year 1986, General Directorate of Turkish Employment Organization has declared that 50 thousand informal workers had been employed in public sector. In 1990, it had been figured out that 7 out of every 10 employers had informal employment. In 1993, it is estimated that more than 2 million workers are informally employed. And by 1999, it was estimated that, the volume of informal employment exceeded 4.5 million (Bildirici:2007). 5 “Non-primary earnings” include earnings from activities other than primary activities of firms (interest, stocks, repurchase agreements etc.). this issue that emerged after 1980 has reached its peak after 1995 and particularly in the years 2000-2001. In the crisis years of 1991, 1994, 1999 and 2001, these earnings reach

96


increase is not observed in national accounts since formal level of production is considered in productivity calculations. Fort his reason, we shall exclude such effects and base the analysis on formal firms and production figures. In the second part of the study, we explain the model and the method we will use while we spare the third part for econometric outcomes. 2 Model In the study, we investigated the TAR unit root and TAR cointegration analysis. In the literature, even though the nonlinear structure and the threshold effects of unemployment are discussed in certain studies, the relationship between productivity and wages do not seem to have been analyzed in accordance with the nonlinear cointegration methods. Rothman (1991); Chen and Lee (1995); Montgomery, Zarnowitz, Tsay, and Tiao (1998); Altissimo and Violante (1996); Chan and Tsay (1998); Hansen (1997); Tsay (1997) analyzed the unemployment accordingly. Further, Caner and Hansen (2001) investigated the unemployment series in accordance with the TAR unit root methodology. 2.1 TAR Unit Root. One point that cannot be overlooked is that, a significant reason of application of the TAR unit root model , as discussed by Caner and Hansen (2001) is the traditional unit root tests (KPSS, ADF, PP etc.) test the hypothesis of stationarity with the nonlinearity also tested within the relevant hypothesis. Thus, it is common that the stationarity is rejected in favor of integration of order d mostly resulting from nonlinearity. On the other hand, Pipenger and Goering (1993) showed that, DF test lost its testing power especially for the TAR models. Tsay (1997) modeled and applied unit root tests but failed to develop the relevant asymptotic distribution theory. Tsay’s model required the usage of constant their highest levels. In the year after the crisis they fall sharply and the next year they rapidly increase one more time. In the years of high “non-primary earnings”, the magnitude of the cost of the employee to the firm becomes unimportant while the growth rate of such earnings decrease (as in the 2001 crisis), firms recourse to practices mentioned above.

97


autoregressive lags. Gonzalez and Gonzalo (1998) investigated the TAR(1) model that has unit roots and tested the threshold in TAR(1). In the study, firstly we will follow the method of Caner and Hansen (2001) and apply the unit root tests. Following the acceptance of the unit root in both regimes for the wage and productivity series for the period investigated, we develop the study followed by the TAR cointegration procedure developed by Hansen and Seo (2002) to investigate the cointegration relation among the variables. Caner and Hansen (2001) put forth that, unemployment is a stationary nonlinear process. Our results confirm the results of Caner and Hansen for the Turkish Economy. Two regime threshold autoregressive model can be written as follows,

∆yt = θ1′xt −1I { zt−1 < λ} + θ 2′ xt −1I {zt −1 ≥ λ} + et where xt−1 = ( yt−1rt′∆yt −1...∆yt −k )′ and I(.) is the indicator function. In model (2.1) errors et : i .i.d . , and Zt = y t − y t− m for some m ≥ 1 . rt is vector of deterministic components. Zt-1 is predetermined, strictly stationary and ergodic with a continuous distribution function. Threshold variable ? is unknown and takes the values in the intervalλ ∈ Λ = [ λ1 , λ2 ] .

In

this

interval

for

?1

and

?2 ,

P( Z t ≤ λ1 )= π1 > 0 , P( Z t ≤ λ2 ) = π 2 < 1 relations can be written. In Caner and Hansen (2001) the components of parameters θ1 and θ 2 are analyzed separately as,  µ1   µ2      θ1 =  β1  , θ 2 =  β2  . ρ  ρ   1  2 In these θ1 and θ 2 vectors ( µ1 , µ 2 ) are slope coefficients,

( β1 , β2 ) ( ρ1 , ρ 2 )

are the slopes on the deterministic components, and are the slope coefficient on yt−1 . For some λ > 2 ,

98


E et

2γ

< ∞ is assumed. For some matrix δ T and continuous vector

function r(s),

δ T r[Ts ] ⇒ r ( s) relation is defined in their analyze.

These assumptions apply that ρ1 = ρ2 = 0 , for constants µ1 and

µ2 , β1′rt = µ1 and β2′ rt = µ 2 , and when i is a vector of ones than a1′i < 1 and a′2i < 1 . For each λ ∈Λ (1) is estimated by OLS,

∆yt = θˆ1( λ )′ xt −1I{ zt−1 < λ} +θˆ2 ( λ )′ xt −1I{ zt−1 ≥ λ} + eˆt ( λ ) OLS estimate of σ 2 for fixed λ , σˆ 2 ( λ ) is, T

2 −1 2 σˆ ( λ ) = T ∑ eˆt ( λ ) 1

and OLS estimate of threshold λ is found by minimizing σ 2 ( λ ) ,

λˆ = argmin σˆ 2 ( λ ) . λ∈Λ

Estimation of the other parameters are done by plugging in λˆ . Then the estimated model is written as,

∆yt = θˆ1′xt−1I

{ zt−1 γ There are two regimes defined by the error correction terms value. As described in Hansen and Seo (2002) the parameters A1 and A2 are coefficient matrices and requires the dynamics in these regimes. If P( wt −1 ≤ γ ) is has the relation 0 < P ( wt −1 ≤ γ ) < 1 this shows threshold effect, otherwise the model characterizes linear cointegration. And also they form the following constraint,

π 0 ≤ P ( wt −1 ≤ γ ) ≤ 1 − π 0 where the trimming parameter π 0 > 0 . 3 Data and Econometric Results The study aims to analyze the relationship between the wages and productivity for 1990:01 – 2007:03 period. To facilitate the relationship between the wages and productivity, indirect taxes and other types of payments are left out of the study; whereas, the real pay received by the worker is kept in the focus of analysis. In order to avoid the deviation in the analysis caused by the unregistered economy, we kept the illegal workers out of the study and we used the registered workers and their contribution to the production. Nonoperating revenues are also kept outside the study. As a result, to avoid the deviation mentioned, ap=the value added of marginal worker and w=wages; w does not include direct/indirect taxes or liabilities. At the second stage, where the unit root procedure suggested by Caner and Hansen (2001) is implemented, we examined the relation between wage and productivity on the basis of Hansen and Seo (2002). As in that approach we also use (MLE) Maximum likelihood estimation of the threshold model. Further, we tested the presence of possible threshold effects, where the model transforms to the linear VECM under the null hypothesis if the alternative is statistically insignificant. 3.1 Unit Root Tests. We applied Augmented Dickey Fuller test to the variables and obtained the lowest AIC information criteria 103


at the 4th and 5th lags. As a result of the test, we failed to reject the hypothesis that the series are stationary and accepted the null hypothesis that the series are integrated of I(1) at the 0,01 significance level. Table 1. Augmented Dickey-Fuller Test W P -0.846160 0.242448 Test critical values: 1% level : -2.576; 5% level: -1.942; 10% level: -1.615

The nonlinear unit root test results are given in Table 1. We based our calculations on Caner and Hansen (2001) methodology. The m parameter for both variables are obtained as 2; whereas k is obtained as 4 for P (productivity) and 5 for W (wage). The lag length k is calculated by AIC information criteria. The m parameter, on the other hand, is calculated in accordance with the Caner and Hansen (2001). 3.2 Bootstrap Threshold Test. The bootstrap threshold test tests the threshold effect on the time series investigated. Consequently, we reported the p values resulting from the Wald test according to the k and m lag lengths obtained. As can be seen in the table, at m lag length both variables contain threshold effect. When the analysis is carried according to k lag length determined by AIC and for whic h we obtained the highest test statistic. As a result, although the variable w has threshold effect at 0,95 confidence level, the H0 hypothesis that points out stationary cannot be accepted for the variable AP at k lag length. However, Caner and Hansen (2001)obtained similar results in their study and accepted to continue the analys is according to m lag length. Consequently, the alternative hypothesis that states threshold effect exists for the series is accepted against the null hypothesis of stationarity. Following the Bootstrap Threshold Test, we analyzed R1 and R2 tests in the study. After the theoretical explanation of R2 test which investigates two sided Wald statistic; the R1 is a one sided test. The test aims to test the null hypothesis of stationarity H0 : ?1 = ?2 = 0 against H1 : ?1 ? ?2 ?0. As observed by the Wald statistics and p values, the null hypothesis of stationarity is rejected for both series by the R2 test for both m and k lag lengths. According to the R1 test we obtained similar results such 104


that the stationarity of the series H0 : ?1 = ?2 = 0 is rejected against the unit root in the series and the alternative hypothesis that series has unit root H1 : ?1 0,83Wt + 0,719 respectively. Another way to interpret the results is that, fist there are two long run cointegration relations for the first and the second regimes defined as the dominant and the extreme for the former and the latter, hence as long as the increase in productivity passes the 0,70 threshold the first regime dominated the economy for the period investigated. References Abuaf, N., and P. Jorion (1990): “Purchasing power parity in the long run,” Journal of Finance, 45, 157-74. Akerlof, G. A. and J. Yellen (1990): “The fair wage-effort hypothesis and unemployement,” Quarterly Journal of Economics, 105, 255-83. Altissimo, F., and G. Violante (1996): ‘‘Persistence and nonlinearity in U.S. GNP and unemployment: an endogenous delay Threshold VAR,’’ Working Paper, University of Pennsylvania. Andrews, D. W. K., and W. Ploberger (1994): “Optimal tests when a nuisance parameter is present only under the alternative,” Econometrica, 62, 1382-414.

107


Balke, N. S., and T. B. Fomby (1997): “Threshold cointegration,” International Economic Review, 38, 627-45. Balke, N.S., and M. E. Wohar (1998): “Nonlinear dynamics and covered interest rate parity,” Empirical Economics, 23, 535–559. Baum, C.F., J. T. Barkoulas, and M. Caglayan (2001): “Nonlinear adjustment to purchasing power parity in the post-Bretton Woods era,” Journal of International Money and Finance, 20, 379–399. Baum, C.F., and M. Karasulu (1998): “Modelling federal reserve discount policy,” Computational Economics, 11, 53–70. Bildirici, M. (2004): “Real cost of employment an Turkish labour market: a panel cointegration tests approach,” International Journal of Applied Econometrics and Quantitative Studies, 1-2, 95-120. Bildirici, M. (2006): “Analysis of Turkish labour market after 1990: panel probit analysis,” YKER, 2006(2) Bildirici, M. (2007): “Türkiye’de Issizlik Sorunsali ve Gaziantep için Sonuçlar,” 14. National Economy Symposium, 25-26 / October / 2007, Gaziantep University Caner, M., and B. E., Hansen (2001): “Treshold autoregressions with a unit root,” Econometrica, 69, 1555-97. Chan K. S. (1993): “Consistency and limiting distribution of the least squares estimator of a threshold autoregressive model,” The Annals of Statistics, Vol. 21, No. 1, 520-33 Chan, K. S. and H. Tong (1985): "On the use of the deterministic Lyapunov function for the ergodicity of stochastic difference equations," Advances in Applied Probability, 17, 666-78. Chan, K. S. and R. S. Tsay (1998): “Limiting properties of the least squares estimator of a continuous threshold autoregressive model,” Biometrika, 85, 413-26. Chen, C. W. S., and J. C. Lee (1995): ‘‘Bayesian Inference of Threshold Autoregressive Models,’’ Journal of Time Series Analysis, 16, 483-92. Davies, R.B. (1987): “Hypothesis testing when a nuisance parameter is present only under the alternative,” Biometrika, 74, 33-43. Delipalla S. and P. Sanfey (2001): "Commodity taxes, wage determination and profits," Journal of Public Economic Theory, 3(2), 203-17. Dickey A. D. and A. W. Fuller (1979): “Distribution of the estimators for autoregressive time series with a unit root ,” Journal of the American Statistical Association, 74, 427-31. Elliott, G., T. J. Rothenberg, and J. H. Stock (1996): “Efficient tests for an autoregressive unit root,” Econometrica, Vol. 64, No. 4, 813-36. Enders, W., and B. Falk (1998): “Threshold-autoregressive, median-unbiased, and cointegration tests of purchasing power parity,” International Journal of Forecasting 14, 171-86. Enders, W., and C.W.J. Granger (1998): “Unit-root tests and asymmetric adjustment with an example using the term structure of interest rates,” Journal of Business and Economic Statistics 16, 304-11. Faria, J. R. (2000): "Supervision and effort in an intertemporal efficiency wage model: the role of the Solow condition," Economics Letters, 67(1), 93-8.

108

Bilidirici, M., Alp, E.A. The Relationships between Wages and Productivity Hansen B. and B. Seo (2002): "Testing for two-regime threshold cointegration in vector error correction models," Journal of Econometrics, 110, 293-318. Hansen, B. E. (1996): “Inference when a nuisance parameter is not identified under the null hypothesis,” Econometrica, 64, 413-30. Hansen, B. E. (1997): "Approximate asymptotic p-values for structural change tests," Journal of Business and Economic Statistics, 15, 60-7. Hansen, B. (2001): "The new econometrics of structural change: dating changes in U.S. labor productivity," Journal of Economic Perspectives, 15, 117-28. Hoel, M. (1990): “Efficiency wages and income taxes,” Journal of Economics, 51, 89-99. Hoon, H. T., and E. S. Phelps (1992): "Macroeconomic shocks in a dynamized model of the natural rate of unemployment," American Economic Review, 82(4), 889-900. Johnson, G.E. and P.R.G. Layard (1986): “The natural rate of unemployement: Explanation and policy,” in O. Ashenfelter and R. Layard, eds., Handbook of labor economics, Vol. II (Amsterdam), 921-99. Kapetanios, G., and Y. Shin (2002): “Unit root tests in three-regime SETAR models,” Working Paper wp 465, Queen Mary, University of London, Department of Economics, Kwiatkowski, D., P. C. B. Phillips, P. Schmidt, and Y. Shin (1992): “Testing the null hypothesis of stationary against the alternative of a unit root,” Journal of Econometrics, 54, 159-78. Lin, CC., and CC. Lai (1994): "The turnover costs and the Solow condition in an efficiency wage model with intertemporal optimization," Economics Letters, 45 (4), 501-5. Lo, M., E. Zivot (2001): “Threshold cointegration and nonlinear adjustment to the law of one price,” Macroeconomic Dynamics 5, 533–76. Lo, M. C. and E. Zivot (2001), “Threshold cointegration and nonlinear adjustment to the law of one price,” Macroeconomic Dynamics, 5(4), 533-76. Malcomson, J. M. and N. Sartor (1987): “Tax push inflation in a unionized labour market,” European Economic Rewiew, 31, 1581-96. Martens, M., P. Kofman, T.C.F. Vorst (1998): “A threshold error-correction model for intraday futures and index returns,”Journal of Applied Econometrics,13, 245–63. Michael, M., R. Nobay, and D. A. Peel (1997): “Transactions costs and nonlinear adjustment in real exchange rates: an empirical investigation,” Journal of Political Economy, 105, 862–79. M ontgomery, A. L., V. ZARNOWITZ, R. S. TSAY, and G. C. TIAO (1998): “Forecasting the U.S. unemployment rate,’’ Journal of the American Statistical Association, 93, 478-93. Newey, W., and K. West (1994): “Automatic lag selection in covariance matrix estimation,” Review of Economic Studies, 61, 631-53. Ng, S., and P. Perron (1995): “Unit root tests in ARMA models with data dependent methods for the selection of the truncation lag,” Journal of the American Statistical Association, 90, 268-81. Ng, S., and P. Peron (2001): “Lag length selection and the construction of unit root tests with good size and power,” Econometrica, 69(6), 1519-54.

109


O’Connell, P.G.J. (1998): “Market frictions and real exchange rates,” Journal of International Money and Finance, 17, 71–95. O’Connell, P.G.J., ans S. J. Wei (1997): “The bigger they are, the harder they fall: how price differences across U.S. cities are arbitraged,” NBER Worker Paper No. 6089. Obstfeld, M., and A.M. Taylor (1997): “Nonlinear aspects of goods market arbitrage and adjustment: Heckscher’s commodity points revisited,” Journal of the Japanese and International Economies, 11, 441–79. Phillips, P. C. B. (1987): “Time series regression with a unit root,” Econometrica, 55, 277-301. Phillips, P.C.B., and P. Perron (1988): “Testing for a unit root in time series regression,” Biometrika, 75, 335-46. Pippenger, M.K., and G. E. Goering (1993): “A note on the empirical power of unit root tests under threshold processes,” Oxford Bulletin of Economics and Statistics, 55, 473-81. Pisauro, G. (1991): “The effect of taxes on labour in efficiency wage models,” Journal of Public Economics, 46, 329-45. Berben, R.P., and D. van Dijk (1999): Unit roots and asymetric adjustment - a reassessment, Econometric Institute Report 101, Erasmus University Rotterdam, Econometric Institute. Rasmussen, B. S. (1998): “Long-run effects of employement and Payroll taxes in an efficiency wage model,” Economics Letters, 58, 245-53. Rothman, P.(1991):“Further evidence on the asymmetric behavior of unemployment rates over the business cycle,” Journal of Macroeconomics, 13, 291-8. Shapiro, C., and J. E. Stiglitz (1984): “Equilibrium unemployment as a worker discipline device,” American Economic Review, 74, 433-44. Taylor, A.M. (2001): “Potential pitfalls for the purchasing-power-parity puzzle? Sampling and specifcation biases in mean-reversion tests of the law of one price,” Econometrica, 69, 473-98. Tong, H. (1978): “On a threshold model,” in C. H. Chen, ed., In Pattern Recognition and Signal Processing, Sijthoff and Noordhoff, Amsterdam, 101-41. Tong, H. (1983): Threshold Models in Non-linear Time Series Analysis. New York, Springer-Verlag. Tong, H. (1990): Non-linear Time Series. A Dynamical System Approach, Oxford: Clarendon Press, Tong, H., and K. S. Lim (1980): “Threshold autoregression, limit cycles and cyclical data,” Journal of Royal Statistical Society, 42(3), 245-92. Tsay, R. S. (1989): “Testing and modeling threshold autoregressive processes,” J. Amer. Statist. Assoc., 84, 231-40. Tsay,R. S. (1997): Unit-root tests with threshold innovations. University of Chicago. Tsay, R. S. (1998): “Testing and modeling multivariate threshold models”. J. Amer. Statist. Assoc., 93, 1188-202. Wu, Y. (1996): “Are real exchange rates nonstationary? Evidence from panel-data tests,” Journal of Money Credit and Banking, 28, 54-63.

Journal published by the EAAEDS: http://www.usc.es/economet/eaa.htm 110